A Flexible Key for Identifying Clinical

Transcription

1 JOURNAL OF CLINICAL MICROBIOLOGY, Dec. 1991, p /91/ $02.00/0 Copyright 1991, American Society for Microbiology Vol. 29, No. 12 Aerokey Il: A Flexible Key for Identifying Clinical Aeromonas Species A. M. CARNAHAN,t S. BEHRAM,t AND S. W. JOSEPH* Department of Microbiology, University of Maryland, College Park, Maryland Received 19 June 1991/Accepted 27 September 1991 A small subset (n = 18) of highly discriminatory tests was derived from the feature frequency of 50 tests used in the study of 167 predominantly clinical Aeromonas strains. Seven of these eighteen tests were used to construct a flexible, dichotomous key, Aerokey II, for identifying clinical aeromonads: esculin hydrolysis, gas from glucose, acid from arabinose, indole production, acid from sucrose, Voges-Proskauer reaction, and resistance to cephalothin (30,ug). This schema was initially evaluated in a single-blind trial of 60 wellcharacterized clinical Aeromonas hydrophila (n = 21), A. caviae (n = 19), and A. veronii bv. sobria (n = 20) strains from an independent laboratory. Of the 60 strains tested, 58 (97%) were accurately identified to the species level. Aerokey Il was further evaluated with 18 additional American Type Culture Collection and reference strains representing the more recently proposed taxa A. veronii bv. veronii, A. schubertii, A. jandaei, and A. trota and accurately identified ail of these strains. During the past 20 years there has been a trend toward the recognition of mesophilic aeromonads as causative agents of human disease (27). Concomitantly, there has been an increase in the numbers of clinical cases reported, originating with sporadic case reports of aquatic wound infections in immunocompromised individuals and expanding to numerous associations with disseminated disease (septicemia, meningitis, and osteomyelitis), gastroenteritis, and wound infections among pediatric and adult populations, both immunocompromised and otherwise healthy (17, 44). This accumulation of information has attracted the interest of not only the medical community but also systematists and infectious disease researchers, as reflected by the publication of four review articles on aeromonads in just the last three years (2, 6, 27, 30). The taxonomy of aeromonads has also been constantly changing. Major taxonomic studies conducted over the last 5 years have provided some clarification of the systematics of aeromonads with respect to the number of DNA hybridization groups (genospecies) and phenotypic species (phenospecies) which currently exist among aeromonads (3, 4, 7, 24, 31). There are now at least seven established or recently proposed taxa of the genus Aeromonas associated with human disease (Table 1). However, often the new additions to the genus Aeromonas have actually contributed to and even exacerbated the existing confusion in the taxonomy. A case in point is the species Aeromonas sobria. Originally proposed by Popoff and Veron in 1976 on the basis of a study of 68 mostly environmental strains, this species was found to encompass at least three separate DNA hybridization groups (DNA groups 7, 8, and 9) with the type strain, CIP 7433, residing in DNA group 7 (14, 39, 40). However, later taxonomic studies of large numbers of clinical isolates revealed that all clinical isolates that resembled the A. sobria phenotype were found by DNA hybridization to actually reside in DNA group 8 (3, 4, 7, 14, 24, 31). * Corresponding author. t Present address: Anne Arundel Medical Center, Annapolis, MD t Present address: Eastern Virginia Medical School, Norfolk, VA Subsequent to this work, a new species, A. veronii (ornithine decarboxylase positive) was proposed as DNA group 10, since it did not hybridize with any of the known Aeromonas type strains, including that of A. sobria (DNA group 7) (22). However, the authors revealed in an addendum in proof to this proposal that clinical A. sobria (DNA group 8) was genetically identical to A. veronii (DNA group 10). The dilemma is that although the species A. sobria (DNA group 7) is indeed a valid one, it is represented by only a small number of environmental strains, including the type strain, CIP To date, all clinical strains resembling A. sobria belong to DNA group 8, not DNA group 7, and since DNA group 8 is genetically identical to DNA group 10 A. veronii, what we have previously identified as clinical A. sobria strains must now be considered as strains of a biovar of A. veronii. We have proposed the names A. veronii bv. veronii for the ornithine decarboxylase-positive strains and A. veronii bv. sobria for those clinical strains that we formerly considered A. sobria in an effort to alleviate the inherent confusion and more accurately reflect what has been discovered at the genospecies level (8). While it may seem confusing to use the epithet sobria as part of the nomenclature for two distinct genospecies, it appears to be at présent the best solution to an already complex situation. Only in this way will clinical microbiologists and physicians be aware that clinical strains formerly identified as A. sobria are now a biovar of A. veronii. It is because of the strong association between these formerly identified A. sobria strains and bacteremia and because of their possession of markers for increased virulence (26, 29) that this change in taxonomic designation must reflect what has already been presented in the literature. Although a majority of publications and most rapid identification systems still use the nomenclature A. hydrophila or A. hydrophila group for any clinical Aeromonas isolate, the tide of taxonomic research clearly suggests that we have the means to identify clinical aeromonads to the species level (3, 7, 9, 25, 39). There is also sufficient evidence to indicate that there are possibly species-related disease syndromes, such as the previously mentioned bacteremia with A. veronii bv. sobria (26), pediatric diarrhea with A. caviae (1, 36), aquatic wound infections with A. schubertii (11, 21), and wound infections with A. hydrophila following the use of medicinal

2 9b Phenotype (phenospecies)a b.a. hydrophila (ATCC 7966T) 2.A. hydrophila 3.A. salmonicida 4b'... A. caviae (ATCC 15468T) 5...A. media 6...A. eucrenophila (ATCC 23309T) 7...A. sobria (CIP 7433T) 8/10b....A. veronii bv. sobriac (ATCC 9071)... A. jandaei (ATCC 49568') 8/1Ob b... A. schubertîi (ATCC 43700T) 2844 CARNAHAN ET AL. TABLE 1. Genotype (DNA group) Current taxonomic status of the genus Aeromonas A. veronii bv. veronii (ATCC 35624T) 11.Aeromonas spp. ornithine decarboxylase positive (ATCC 35941) 13.Aeromonas group 501 (ATCC 43946) 14b... A. trota (ATCC 49657T) a Strains in parentheses are type (T) or representative strains. b Genospecies isolated from clinical specimens (3, 4, 8, 10, 21, 22, 24). c Formerly identified as clinical A. sobria, but genetically identical to A. veronii (3, 8, 22). leeches (42). This variability in pathogenicity, coupled with the existence of different antibiograms among the species, e.g., resistance to cephalothin (28) or susceptibility to ampicillin (10), could affect a physician's subsequent successful treatment of a patient. However, no previous studies have resulted in a workable identification schema for the mesophilic Aeromonas species associated with human disease. Hence, most clinical isolates are still identified as either A. hydrophila or A. hydrophila group whether conventional methods or rapid identification systems are used. The objective of this research was to examine a large number of clinical Aeromonas isolates from diverse clinical and geographic sources and use the frequency matrix of test results from a numerical taxonomy analysis to develop a highly discriminatory subset of tests. These tests (n = 18) were then used to construct a flexible dichotomous identification key, Aerokey (proprietary technology), that with only seven tests allows a clinical laboratory to accurately identify Aeromonas isolates to the species level. (Part of this work was conducted by A. M. Carnahan in partial fulfillment of the requirements for an M.S. from the University of Maryland, College Park, and was subsequently presented at the 3rd International Aeromonas/Plesiomonas Symposium, Helsing0r, Denmark, 5 to 6 September 1990.) MATERIALS AND METHODS Bacterial strains. One hundred fifty-two strains, initially received from 1985 to 1989 as Aeromonas spp., were used for this study. They were collected in several geographical regions, including the United States (n 78), northeastern = Africa (Somalia, the Sudan, and Egypt) (n 50), Bangladesh = (n = 19), India (n = 1), Indonesia (n 3), and Puerto Rico = (n = 1). The majority of the strains came from a variety of clinical sources (n 131), but = some were veterinary (n 8) = or environmental (n = 13) strains. All major American Type Culture Collection (ATCC) type strains, the Centers for Disease Control reference definition strains for DNA hybridization groups 1, 2, 3, 4, 5a, 5b, 6, 7, 8X, 9, 10, 11, and 12, and ATCC reference strains for A. veronii bv. sobria (DNA J. CLIN. MICROBIOL. group 8Y) and Aeromonas group 501 (DNA group 13) were included as controls. The strains were stored at -70 C in Trypticase soy broth with 10% glycerol (Remel, Lenexa, Kans.) and subsequently subcultured to tryptic soy agar (TSA) plates (Becton Dickinson Microbiology Systems, Cockeysville, Md.) and incubated overnight at 36 C. Unless stated otherwise, all analyses were performed at 36 ± 1 C. Presumptive identification. All strains were initially screened by using the following tests: Gram stain, oxidase activity (1% solution of p-aminodimethylaniline oxalate; Difco, Detroit, Mich.), glucose fermentation using a triple sugar iron (TSI) slant, motility by "wet mount" after incubation at 25 C, and resistance to 0/129 (a vibriostatic agent) (150,ug/ml; Oxoid, Ogdensburg, N.Y.). Only those strains that were motile, oxidase positive, glucose-fermenting, 0/129-resistant, gram-negative rods were considered aeromonads. They were then presumptively identified as belonging to the A. hydrophila group with an API-20E strip (Analytab, Inc., Plainview, N.Y.). Biochemical phenotypic markers. Each strain was further examined for production of diffusible pigments on TSA; indole production (both Kovacs's and Ehrlich's methods); esculin hydrolysis; growth in KCN; acid production from arabinose, salicin, mannitol, and sucrose; production of gas from glucose; production of ornithine decarboxylase; String test (20); arbutin hydrolysis (16); and H2S production from cysteine with a modified gelatin-cysteine-thiosulfate (GCF) medium (43). Additional phenotypic tests that have been associated with specific biotypes or used as potential virulence-associated markers were also evaluated. These were lysine decarboxylase production, the Voges-Proskauer reaction, hemolysis on TSA plates with 5% sheep blood by streaking and stabbing, hemolysis on TSA plates with 5% horse blood by streaking and stabbing, acriflavine agglutination (32), autoagglutination (29), production of a CAMP-like factor (aerobically and anaerobically) (15), cephalothin susceptibility (28), and pyrazinamidase activity (9). Excluding the tests on the API-20E strip and those referenced above, all tests were performed by conventional methods (13, 34) based on the Janda modification (25) of the original schema of Popoff and Veron (39) with media and reagents supplied by Remel. Both positive and negative control Aeromonas ATCC strains were included for each test (Table 1, footnote b). Comparison of test methods. Several of the tests (n = 11) used in the initial phenotypic analysis of these strains were performed by two or sometimes three different methods, or with different media formulations, in an effort to determine whether the results were comparable to each other. These results are listed in Table 2 as percent positive feature frequencies (percentage of strains giving positive results) for the different methods or media used. Antibiotic resistance markers. MICs were determined for 17 antimicrobial agents in a 96-well microdilution plate (Gram Negative Panel 7; MicroScan Division, Baxter Healthcare Corp., West Sacramento, Calif.) per the manufacturer's instructions. In addition to these 17 antimicrobial agents, the MicroScan panel tested for a "growth" or "no growth" response against the antimicrobial agents colistin, nitrofurantoin, kanamycin, and penicillin in single-dilution wells. The MIC results were recorded both manually and with an AutoScan-4 reader with computer-assisted analysis by an IBM PS/2 Model 60 which provided interpretations in accordance with the guidelines of the National Committee for Clinical Laboratory Standards (NCCLS) (37). The antimicro-

3 VOL. 29, 1991 AEROKEY Il FOR IDENTIFYING AEROMONAS SPECIES 2845 bial agents whose MICs were determined were amikacin, ampicillin, cefazolin, cefoxitin, ceftazidime, ceftriaxone, cefuroxime, cephalothin, ciprofloxacin, gentamicin, imipenem, piperacillin, tetracycline, ticarcillin-potassium clavulanate, ticarcillin, tobramycin, and trimethoprim-sulfamethoxazole. Cephalothin susceptibility was also tested by the Bauer- Kirby agar disk diffusion method (5) using a 30-,ug disk and Mueller-Hinton agar (Becton Dickinson Microbiology Systems) per NCCLS standards. Numerical taxonomy analysis. Examination of the percent positive frequency results for all of the strains examined revealed those phenotypic tests, both biochemical and antimicrobial, whose results varied from strain to strain at a level greater than 5% but less than 95%. Consequently, only the results for those 50 tests were used in the numerical taxonomy analysis. The data were examined with the SAS/ TAXAN (SAS Institute, Cary, N.C.) clustering program (Maryland Sea Grant College, College Park) on an IBM 4381 mainframe computer by using the simple matching (SsM) coefficient at a similarity level of 85% to delineate clusters of strains that represented distinct phenospecies (7). Only those clusters which represented the established and proposed clinical species were used to construct Aerokey II, and each cluster contained the appropriate ATCC or DNA definition group strain for that phenospecies. These clusters were composed of A. hydrophila (n = 46), A. caviae (n = 33), A. veronii bv. sobria (n = 26), A. veronii bv. veronii (n = 4), A. schubertii (n = 3), A. jandaei (n = 4) and A. trota (n = 7). Once the phenotypic clusters were determined, a frequency matrix was constructed from all of the test results (percent positive results of each of the clusters) on an IBM PS/2 Model 50 computer using a FREQ program written in BASIC by one of the authors of this article (S. Behram). Next, the data base was searched for those tests whose frequencies of positive results were above 70% or below 30% for these specified clusters. Only these tests were used in constructing the identification keys. FLOABN, a greedy algorithm program. Another computer program, FLOABN (For Lack of a Better Name), which used the frequency data from the SAS/TAXAN analysis, was written in BASIC by S. Behram as a modified version of a recently published Pascal program (33). The FLOABN (proprietary technology) program compared the SAS/ TAXAN-derived clusters for the clinical phenotypic species and, using the frequency matrix of results outlined above, constructed a dichotomous identification key by solving a simple test selection problem with a "greedy algorithm." FLOABN was written with a semiautomatic mode that allowed the operator, while constructing the dichotomous key, to select at each level from a number of equivalent, first-choice tests based on media and methods available in clinical laboratories. For this study, several different runs were conducted with many different equivalent first-choice tests at each level, and the result was a small subset (n = 18) of highly discriminatory tests (Table 3). Aerokey I. The clusters initially examined by FLOABN were those that encompassed all of the DNA hybridization groups and ATCC type and reference strains (n = 17). This resulted in a dichotomous key, Aerokey 1 (7), that uses 12 of the original 50 tests to separate all currently recognized genospecies (DNA hybridization groups) except DNA groups 2 and 3 (data not shown). Because the collection of strains often had only a single strain to represent the rare environmental genospecies, e.g., A. eucrenophila and A. media (7), Aerokey I should not yet be considered statistically valid for the identification of every genospecies. Indole Production A. hubertii A. tro Voges Proska Acid f Sucrose A. jand i Oxidase-positive, Glucose-fermenting, Gram negative rods, resistant to 0/129 Vibriostatic Agent (150 g/g) M- muer Erom +1 l 1 A veron i biovar sobna 1 b l Esculin Hydrolysisa A+ Gas from Glucose (TSI) _+ A. caviae Acid from Arabinose Resistance to Cephalothinc (30 &g) _,_ài + A. hydrophila A. veron A. hrophila biovar veroni FIG. 1. Aerokey Il identification key for clinical Aeromonas species. a, agar formulation only; b, Aerokey Il can be modified to end here with an identification of A. veronii bv. sobria; c, Bauer- Kirby disk diffusion method only. Rather, it was constructed to serve as an experimental guideline for those laboratories attempting to identify large groups of aeromonads from clinical, veterinary, and environmental sources for taxonomic, epidemiological, or virulence-related studies (unpublished data). Aerokey Il. Subsequently, the FLOABN program was used to examine those clusters that represented all seven clinical Aeromonas taxa: A. hydrophila, A. caviae, A. veronii bv. sobria, A. veronii bv. veronii, A. schubertii, A. jandaei, and A. trota. This process generated Aerokey II, which contains seven tests for identification to the species level (Fig. 1). The validity of Aerokey Il was first evaluated for the identification of the three mesophilic Aeromonas species most commonly encountered in clinical laboratories. This was accomplished through a single-blind trial of 60 arbitrarily chosen clinical Aeromonas isolates from the private collection of an independent laboratory (Table 4). The strains had already been extensively phenotyped, but for this study they were identified by code number only and accompanied by the results for the seven tests used in Aerokey Il. These tests were for esculin hydrolysis, gas from glucose, acid from arabinose, indole production, acid from sucrose, the Voges-Proskauer reaction, and resistance to cephalothin 1

4 2846 CARNAHAN ET AL. TABLE 2. Results of comparative test methods and media Test and medium Esculin hydrolysis (EHA)... Bile esculin hydrolysis (BEHA)... Gas from glucose TSI... 1% Andrade's broth with Durham vial... Cephalothin resistance Bauer-Kirby method... MicroScan MIC method... Lysine decarboxylase Moeller's method... Ornithine decarboxylase Moeller's method... Indole production Kovacs's reagent on TSA with 5% sheep blood... Voges-Proskauer MR-VP' broth... Acid from arabinose 1% Andrade's broth... Acid from mannitol 1% Andrade's broth... Acid from sucrose 1% Andrade's broth... H2S Production Modified GCF medium... TSI agar stab... bno. (%) of strains testing positive (n 167) 114 (68) 114 (68) 91 (54) 90 (54) 119 (71) 122 (73) 102 (61) 97 (58) 11 (7) 5 (3) 158 (95) 156 (93) 95 (57) 87 (52) 92 (55) 96 (57) 156 (93) 159 (95) 145 (87) 146 (87) 102 (61) 5 (3) O (0) a EHA, esculin hydrolysis agar; BEHA, bile esculin hydrolysis agar. MR-VP, methyl red-voges-proskauer. (Fig. 1). Acid from arabinose is present at two branches in the key because of the possible existence of two distinct biovars of A. hydrophila within the same genospecies (7). Because of the rare isolation to date of the more recently proposed taxa A. veronii bv. veronii, A. schubertùi, A.jandaei, and A. trota, 18 additional ATCC and reference strains for these species were obtained and tested with Aerokey Il. RESULTS Comparison of test methods and media. The ability of aeromonads to hydrolyze esculin was evaluated with both esculin hydrolysis agar and bile esculin hydrolysis agar. Both media were incubated at 36 C for 72 h. This was done to determine the possibility of an inhibitory effect on the growth of aeromonads by the bile in bile esculin hydrolysis agar, which would preclude a positive test result on this J. CLIN. MICROBIOL. medium. Bile esculin hydrolysis agar was found to be equivalent to esculin hydrolysis agar in sensitivity and specificity, with no false positives or false negatives occurring on the bile formulation medium (Table 2). It can be concluded that bile esculin hydrolysis agar, available in most clinical laboratories for the identification of Enterococcus spp., can be used for identifying aeromonads. However, it should be noted that the evaluation of two different rapid identification systems' esculin broth formulation, using these same strains, resulted in a large number of false-negative reactions (no black precipitate), and therefore the broth formulation is not considered comparable to the agar formulation pending further studies (7). The production of gas from glucose fermentation was evaluated by examination of an 18-h TSI slant as well as the conventional tube method using Andrade's broth with 1% dextrose and a Durham vial for detection of gas. The TSI slant and butt results were determined after 18 to 24 h at 36 C, while the Andrade's broth was incubated for 72 h at 36 C. All but one of the 167 strains tested yielded identical results in the two tests, with one strain being considered positive with TSI only, suggesting that the use of TSI for gas production is as sensitive as Andrade's broth with a Durham vial. Susceptibility to cephalothin was determined by both the MIC dilution method (MicroScan) and the Bauer-Kirby agar disk diffusion method, and the only discrepancy was that three strains shown by the MIC method to be resistant to cephalothin were shown by the disk diffusion method to be susceptible. However, if resistance to cephalothin is to be used in Aerokey Il as a differential test, only the standard Bauer-Kirby method should be employed pending further evaluation. This recommendation is also made because of the possibility of other major discrepancies in the detection of cephalothin resistance by inducible P-lactamases of Aeromonas spp. when an MIC microdilution method (Vitek) is used (41). The following tests were performed both by conventional methods and with an API-20E strip: lysine and ornithine decarboxylase (API-20E versus Moeller's method); indole production (API-20E versus the conventional spot indole test on a 24-h culture grown on TSA with 5% sheep blood and Kovacs's reagent, but with all negatives confirmed by a conventional tryptone broth with xylene extraction and Ehrlich's reagent); Voges-Proskauer test (API-20E versus methyl red-voges-proskauer broth with the modified O'Meara method); and acid from arabinose, mannitol, and sucrose (API-20E versus Andrade's broth with 1% carbohydrate). The results for the two decarboxylase enzymes were in close agreement for the two different methods, as were the results for indole, the Voges-Proskauer test, and carbohydrate fermentation (Table 2). The API-20E formulation was often more sensitive, e.g., in the case of ornithine decarboxylation. This suggests that results obtained in these tests with API-20E strips (per the manufacturer's instructions) correlate very well with the results obtainable with conventional test media for identifying aeromonad species. Because production of H2S by aeromonads has been a subject of controversy (18, 25, 31, 39), three different methods were evaluated: a 72-h TSI agar stab, a 24-h H2S test on an API-20E strip (per the manufacturer's instructions), and a test using semisolid medium, GCF medium. This medium, a modification of an original Veron and Gasser medium, was inoculated by stabbing and then was incubated for 72 h at C and read at 24-h intervals. (43). The differences in the percentage of strains positive by these three test methods were readily apparent, in that none of the 167 strains gave a

5 VOL. 29, 1991 AEROKEY Il FOR IDENTIFYING AEROMONAS SPECIES 2847 TABLE 3. Comparison of distinguishing profiles of mesophilic clinical Aeromonas species Resulta for: Characteristic A. A. hydrophil. A. veronii bv. A. veronii bv. A. caviae A. schubertii A. jandaei A. trota 46)ph la sobria veronii A = = (n=46) 33) (n-6) An ~~(n =26) (n 9.tno13 = 8) (n=) (n6)n=9 (1) Esculin hydrolysis Voges-Proskauer reaction V + Pyrazinamidase activity CAMP-like factor (aerobic V only) Fermentation Arabinose V Mannitol Sucrose Susceptibility Ampicillin R R R R R R S Carbenicillin R R R R R R S Cephalothin R S S R S R R Colistinb V S S S S R S Decarboxylase Lysine Ornithine - + Arbutin hydrolysis V Indole H2SC Glucose (gas) Hemolysis (TSA with 5% V + + V sheep erythrocytes) a +, positive for >70% of isolates; -, negative, i.e., positive for <30% of isolates; V, variable; R, resistant, S, susceptible. b MIC (single dilution), 4,ug/ml. ' H2S from GCF medium. d Modified from Table 4 of reference 10 with permission. positive result for H2S production with the API-20E strip and only 5 (3%) gave a positive result with the TSI slant (Table 2). With the modified GCF medium, 102 (61%) of the 167 strains were positive for H2S production. Further, the results obtained supported the H2S production test as one that can differentiate between the species, as originally stated by Popoffand Veron (39) and later supported by Janda (25). That is, all of the SAS/TAXAN-phenotyped A. hydrophila and A. veronii bv. sobria strains were positive (100%) for H2S, while all of the A. caviae strains were negative. Therefore, production of H2S can be a useful test for identifying clinical aeromonads if modified GCF medium is used. Another controversial test is growth in KCN broth as first proposed by Popoff and Veron (39). Although our results confirmed those of the initial study (39) as well as the later work of Janda (25), we found this test to be very subjective in its interpretation and both hazardous and labor intensive in its preparation (7). The inherent discrepancies associated with the interpretation of growth in KCN broth and the detection of H2S production may explain, in part, the variance of our results compared with those of George et al. (18) and Kuijper et al. (31) concerning the identification of clinical Aeromonas species. Finally, the method of Figura and Guglielmetti for the CAMP-like test for the identification of Aeromonas species was evaluated (15). We found that only the aerobic CAMPlike test could accurately differentiate A. hydrophila and A. veronii bv. sobria from A. caviae (Table 3). In the anaerobic case, we often observed a positive CAMP-like "arrow" with both A. hydrophila and A. veronii bv. sobria, not just A. hydrophila as originally stated (15). This variance between our findings and those originally published may be due to the fact that the original work was conducted with only a small number of strains. Also, a taxonomic study of a large number of clinical strains by Altwegg et al. found 16% of the A. veronii bv. sobria strains tested to be positive when the CAMP-like test was conducted anaerobically (3). Evaluation of Aerokey Il. When Aerokey Il was used to identify the 60 coded clinical strains received from the independent laboratory, 58 of 60 (97%) of the strains were correctly identified to the species level. Only two highly atypical A. veronii bv. sobria (esculin hydrolysis positive and cephalothin resistant) strains were incorrectly identified as A. hydrophila (Table 4). It is interesting to note that one strain identified as A. jandaei by Aerokey Il had been TABLE 4. Evaluations of Aerokey Il Phenospecies No. of strains No. tested identified (%) correctly Clinical strains (single-blind trial) (n = 60)' A. hydrophila (100) A. caviae (100) A. veronii bv. sobria (90) A. jandaei 1 1 (100) ATCC or DNA reference strains (n = 18) A. veronii bv. veronii 4 4 (100) A. schubertii 3 3 (100) A. jandaei 5 5 (100) A. trota 6 6 (100) a Sources of clinical strains: feces (n = 31); wounds (n = 12); blood (n = 10); bile (n = 3); sputum, nares, urine, and cerebrospinal fluid (n = 1 each).

6 2848 CARNAHAN ET AL. originally phenotyped as A. veronii bv. sobria but was subsequently confirmed by DNA-DNA hybridization to indeed be A. jandaei (formerly DNA group 9 A. sobria) (8). The evaluation of Aerokey Il with 18 additional ATCC or DNA-hybridized reference strains of the more recently proposed taxa A. veronii bv. veronii, A. schubertii, A. jandaei, and A. trota resulted in all 18 (100%) being correctly identified to the species level (Table 4). DISCUSSION While it is quite clear that aeromonads can cause a myriad of infections in human hosts, controversy as to whether we can identify clinical, mesophilic Aeromonas isolates to the species level still exists. Some researchers express concern that there is not a sufficient quantity of discriminatory tests to distinguish between the clinical species or that the tests available are not sufficiently accurate (18, 22, 24). However, taxonomic studies of large numbers of clinical isolates have repeatedly shown that the majority of clinical Aeromonas isolates reside in DNA groups 1, 4, and 8, which are phenotypically characterized as A. hydrophila, A. caviae, and A. veronii bv. sobria (2, 4, 7, 24, 31). Additionally, biochemical studies of large numbers of Aeromonas isolates have provided a number of tests that can clearly separate the clinically significant species (3, 4, 7, 9, 25, 28, 39). The more recently proposed taxa A. veronii bv. veronii, A. schubertii, A. jandaei, and A. trota, although isolated in smaller numbers thus far, do have unusual biochemical markers that facilitate screening for these taxa in clinical laboratories (8, 10, 21, 22). However, there was no single identification schema for identifying clinical Aeromonas isolates to the species level that incorporated all of these findings. The previous taxonomic studies of large numbers of clinical aeromonads focused almost exclusively on fecal isolates, usually from just one geographic location (3, 4, 24, 31), whereas our research encompassed a variety of clinical isolates from very diverse geographic locations with a smaller number of veterinary and environmental isolates as well. It was only when a large number of diverse strains were examined by an extensive number of tests using standard methods of identification with well-defined media and reagents that a workable identification system was developed. When Aerokey Il is implemented in many different laboratories, we can begin to measure the true frequency and distribution of Aeromonas species from clinical isolates both within and between laboratories. The ability of Aerokey Il to correctly identify 97% of the 60 coded clinical isolates from an independent laboratory and 100% of the reference strains to the species level qualifies Aerokey II as a reliable and accurate system for the identification of the Aeromonas taxa isolated from clinical specimens to date. Additionally, use of the other discriminatory tests listed in Table 3 in conjunction with Aerokey Il allows microbiologists to substitute tests at each step as desired; e.g., resistance to cephalothin could be replaced with either pyrazinamidase activity or ornithine decarboxylase to differentiate between A. veronii bv. veronii and A. hydrophila. Finally, Aerokey II can be modified to end on the left side of the dichotomous key with a positive reaction for indole production leading to an identification of A. veronii bv. sobria (dashed line in Fig. 1), should a clinical laboratory choose not to screen for the newer species at this time. Aerokey Il has already been incorporated at the bench J. CLIN. MICROBIOL. level of the hospital laboratory of one of the authors of this article (A. M. Carnahan) and involves setting up the following for each Aeromonas isolate: an API-20E strip, a TSI slant, a bile esculin hydrolysis agar slant, an 0/129 disk (150 p.g/ml), and a Bauer-Kirby antimicrobial susceptibility panel that includes cephalothin (30,ug). Because of the good correlation between results derived by conventional methods and those derived by the API-20E strip method (Table 2), four ofthe seven tests needed for Aerokey Il can be taken from the API-20E strip: Voges-Proskauer reaction, acid from arabinose, acid from sucrose, and indole production (negative indole reactions should be confirmed by Ehrlich's method). However, it is imperative that all isolates be presumptively identified as Aeromonas spp. before Aerokey Il is used, i.e., they must be oxidase-positive, glucosefermenting, gram-negative rods that are resistant to 0/129 (150,ug/ml). Aeromonads are also unable to grow in a 6.5% NaCI broth. It is equally important that all tests based on Aerokey Il are set up from a pure overnight culture of the Aeromonas isolate taken from a single colony. This procedure is recommended because biochemical tests set up directly from a colony growing on an inhibitory primary plate such as cefsulodin irgasan novobiocin agar (CIN) or MacConkey agar may carry over minute colonies of organisms such as Enterococcus spp. that would give a falsepositive result for esculin hydrolysis at the start of Aerokey Il. Finally, we reiterate that the esculin hydrolysis test must be set up with an agar formulation only and that the determination of resistance to cephalothin must be accomplished by the Bauer-Kirby method only. Because of the absence of large outbreaks of disease caused by Aeromonas spp. and a single, unsuccessful human volunteer trial, we have thus far been unable to consider Aeromonas spp. definitively as singular, significant causative agents of human gastroenteritis (35). Although Koch's postulates have not been fulfilled, promising research on animal models (19, 38) and virulence features such as hemolysin and pili (12, 23) is being done. Perhaps, when accurate identification to the species level is combined with ongoing and future studies on pathogenesis and epidemiology, we can begin to determine why only certain genospecies appear to predominate in clinical specimens, whether they are equally distributed in the environment, and most importantly what is distinguishable about those subsets of aeromonads within each species that appear to be capable of causing disease in humans. Such information will be invaluable in understanding the role of the immune status of the host and, perhaps, culminate in successful human volunteer trials. ACKNOWLEDGMENTS We thank J. J. Farmer lil for the donation of the DNA hybridization group definition strains and J. Michael Janda for the 60 confirmed clinical Aeromonas isolates. We acknowledge the supplies provided by Lisette Gonzalez (API; Analytab Products) and MicroScan and thank Afsar Ali and Marie Head for technical assistance. We also value the critical review of the manuscript by Donna Watsky. This work was done with the support of Agency for International Development grant DPE-5542-G-SS REFERENCES 1. Altwegg, M Aeromonas caviae: an enteric pathogen? Infection 13: Atlwegg, M., and H. K. Geiss Aeromonas as a human pathogen. Crit. Rev. Microbiol. 16: Altwegg, M., A. G. Steigerwalt, R. Altwegg-Bissig, J. Luthy- Hottenstein, and D. J. Brenner Biochemical identification

7 VOL. 29, 1991 AEROKEY Il FOR IDENTIFYING AEROMONAS SPECIES 2849 of Aeromonas genospecies isolated from humans. J. Clin. Microbiol. 28: Arduino, M. J., F. W. Hickman-Brenner, and J. J. Farmer III Phenotypic analysis of 132 Aeromonas strains representing 12 DNA hybridization groups. J. Diarrhoeal Dis. Res. 6: Bauer, A. W., W. M. M. Kirby, J. C. Sherris, and M. Turck Antibiotic susceptibility testing by a standardized single disk method. Am. J. Clin. Pathol. 45: Cahill, M. M A review: virulence factors in motile Aeromonas species. J. Apple. Bacteriol. 69: Carnahan, A., S. Behram, A. Ali, D. Jacobs, and S. W. Joseph Systematic assessment of geographically diverse Aeromonas spp. as a correlate to accurate biotyping of clinical aeromonads, abstr. R-15, p Abstr. 90th Annu. Meet. Am. Soc. Microbiol American Society for Microbiology, Washington, D.C. 8. Carnahan, A., G. R. Fanning, and S. W. Joseph Aeromonas jandaei (formerly genospecies DNA group 9 A. sobria), a new sucrose-negative species isolated from clinical specimens. J. Clin. Microbiol. 29: Carnahan, A., L. Hammontree, L. Bourgeois, and S. W. Joseph Pyrazinamidase activity as a phenotypic marker for several Aeromonas spp. isolated from clinical specimens. J. Clin. Microbiol. 28: Carnahan, A. M., T. Chakraborty, G. R. Fanning, D. Verma, A. Ali, J. M. Janda, and S. W. Joseph Aeromonas trota sp. nov., an ampicillin-susceptible species isolated from clinical specimens. J. Clin. Microbiol. 29: Carnahan, A. M., M. A. Marii, G. R. Fanning, M. A. Pass, and S. W. Joseph Characterization of Aeromonas schubertii strains recently isolated from traumatic wound infections. J. Clin. Microbiol. 27: Chakraborty, T., A. Schmid, S. Notermans, and R. Benz Aerolysin of Aeromonas sobria: evidence for formation of ion-permeable channels and comparison with alpha-toxin of Staphylococcus aureus. Infect. Immun. 58: Edwards, P. R., and W. H. Ewing Identification of Enterobacteriaceae, 3rd ed. Burgess Publishing Co., Minneapolis. 14. Fanning, G. R., F. W. Hickman-Brenner, J. J. Farmer III, and D. J. Brenner DNA relatedness and phenotypic analysis ofthe genus Aeromonas, abstr. C 116, p Abstr. 85th Annu. Meet. Am. Soc. Microbiol American Society for Microbiology, Washington, D.C. 15. Figura, N., and P. Guglielmetti Differentiation of motile and mesophilic Aeromonas strains into species by testing for a CAMP-like factor. J. Clin. Microbiol. 25: Frank, S. K., and L. von Rieson Aglycone tests determine hydrolysis of arbutin, esculin, and salicin by nonfermentative gram-negative bacteria. Lab. Med. 9: Frei, B. J Aeromonas: biology of the organism and disease in children. Pediatr. Infect. Dis. J. 3: George, W. L., M. J. Jones, and M. M. Nakata Phenotypic characteristics of Aeromonas species isolated from adult humans. J. Clin. Microbiol. 23: Haberberger, R. L., W. P. Yonushonis, R. L. Daise, I. A. Mikhail, and E. A. Ishak Re-examination of Rattus norvegicus as an animal model for Aeromonas-associated gastroenteritis in man. Experientia 47: Hickman, F. W., J. J. Farmer III, D. G. Hollis, G. R. Fanning, A. G. Steigerwalt, R. E. Weaver, and D. J. Brenner Identification of Vibrio hollisae sp. nov. from patients with diarrhea. J. Clin. Microbiol. 15: Hickman-Brenner, F. W., G. R. Fanning, M. J. Arduino, D. J. Brenner, and J. J. Farmer III Aeromonas schubertài, a new mannitol-negative species found in human clinical specimens. J. Clin. MIcrobiol. 26: Hickman-Brenner, F. W., K. L. MacDonald, A. G. Steigerwalt, G. R. Fanning, D. J. Brenner, and J. J. Farmer III Aeromonas veronii, a new ornithine decarboxylase-positive species that may cause diarrhea. J. Clin. Microbiol. 25: Ho, A. S. Y., T. A. Mietzner, A. J. Smith, and G. K. Schoolnik The pili of Aeromonas hydrophila: identification of an environmentally regulated " mini pilin." J. Exp. Med. 172: Holmberg, S. D., W. L. Schell, G. R. Fanning, I. K. Wachsmuth, F. W. Hickinan-Brenner, P. A. Blake, D. J. Brenner, and J. J. Farmer III Aeromonas intestinal infections in the United States. Ann. Intern. Med. 105: Janda, J. M Biochemical and exoenzymatic properties of Aeromonas species. Diagn. Microbiol. Infect. Dis. 3: Janda, J. M., and R. Brenden Importance of Aeromonas sobria in Aeromonas bacteremia. J. Infect. Dis. 155: Janda, J. M., and P. S. Duffey Mesophilic aeromonads in human disease: current taxonomy, laboratory identification and infectious disease spectrum. Rev. Infect. Dis. 10: Janda, J. M., and M. R. Motyl Cephalothin susceptibility as a potential marker for the Aeromonas sobria group. J. Clin. Microbiol. 22: Janda, J. M., L. S. Oshiro, S. L. Abbott, and P. S. Duffey Virulence markers of mesophilic aeromonads: association of the autoagglutination phenomenon with mouse pathogenicity and the presence of a peripheral cell-associated layer. Infect. Immun. 55: Khardori, N., and V. Fainstein Aeromonas and Plesiomonas as etiological agents. Annu. Rev. Microbiol. 42: Kujper, E. J., A. G. Steigerwaht, B. S. C. I. M. Schoenmakers, M. F. Peeters, H. C. Zanen, and D. J. Brenner Phenotypic characterization and DNA relatedness in human fecal isolates of Aeromonas spp. J. Clin. Microbiol. 27: Lallier, R., K. R. Mittal, D. Leblanc, G. Lalonde, and G. Olivier Rapid methods for the differentiation of virulent and nonvirulent Aeromonas hydrophila strains. Dev. Biol. Strand. 49: Matousek, J., and J. Schindler Selecting a small welldiscriminating subset of tests from a frequency matrix. Binary 1: McFaddin, J. F Biochemical tests for identification of medical bacteria, 2nd ed. The Williams & Wilkins Co. Baltimore. 35. Morgan, D. R., P. C. Johnson, H. L. DuPont, T. K. Satterwhite, and L. V. Wood Lack of correlation between known virulence properties of Aeromonas hydrophila and enteropathogenicity for humans. Infect. Immun. 50: Namdari, H., and E. J. Bottone Microbiologic and clinical evidence supporting the role ofaeromonas caviae as a pediatric enteric pathogen. J. Clin. Microbiol. 28: National Committee for Clinical Laboratory Standards Methods for dilution antimicrobial susceptibility tests for bacteria that grow aerobically. Approved standards M7-A and M100-2S. National Committee for Clinical Laboratory Standards, Villanova, Pa. 38. Pazzaglia, G., R. B. Sack, A. L. Bourgeois, J. Froehlich, and J. Eckstein Diarrhea and intestinal invasiveness of Aeromonas strains in the removable intestinal tie rabbit model. Infect. Immun. 58: Popoff, M., and M. Veron A taxonomic study of the Aeromonas hydrophila-aeromonas punctata group. J. Gen. Microbiol. 94: Popoff, M. Y., C. Coynault, M. Kiredjian, and M. Lemelin Polynucleotide sequence relatedness among motile Aeromonas species. Curr. Microbiol. 5: Schadow, K. H., D. K. Giger, and C. C. Sanders The Vitek system fails to detect P-lactam resistance in Aeromonas, abstr. 171, p Abstr. Program 31st Intersci. Conf. Antimicrob. Agents Chemother. American Society for Microbiology, Washington, D.C. 42. Snower, D. P., C. Ruef, A. P. Kuritza, and S. C. Edberg Aeromonas hydrophila infection associated with the use of medicinal leeches. J. Clin. Microbiol. 27: Veron, M., and F. Gasser Sur la détection de l'hydrogène sulfuré produit par certaines enterobacteriacées dans les milieux de diagnostic rapide. Ann. Inst. Pasteur 105: von Graevenitz, A., and A. H. Mensch The genus Aeromonas in human bacteriology. N. Engl. J. Med. 278: